ocr_engine.py

import cv2
import numpy as np
import re
from PIL import Image

def extract_weight_from_image(pil_img):
    try:
        img = np.array(pil_img)

        # Convert to grayscale
        gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)

        # Threshold image
        _, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

        # Invert if needed
        if np.mean(thresh > 127) < 0.5:
            thresh = cv2.bitwise_not(thresh)

        # Resize to make digits bigger
        scale_factor = 4
        resized = cv2.resize(thresh, None, fx=scale_factor, fy=scale_factor, interpolation=cv2.INTER_LINEAR)

        # OCR-style region crop: focus on left part of display
        height, width = resized.shape
        digit_region = resized[0:height, 0:int(width * 0.7)]  # ignore 'kg'

        # Use pytesseract as fallback OCR for just digits
        import pytesseract
        config = "--psm 7 -c tessedit_char_whitelist=0123456789."
        result = pytesseract.image_to_string(digit_region, config=config)
        print("Raw OCR:", result)

        match = re.search(r"(\d{1,4}(?:\.\d{1,2})?)", result)
        if match:
            return f"{match.group()} kg", 100.0
        else:
            return "No weight detected kg", 0.0

    except Exception as e:
        return f"Error: {str(e)}", 0.0